Semantic Mapping
Overview
Semantic mapping links observations or data fields to meaningful concepts, categories, entities, and relations so machines can reason beyond raw measurements. In robotics, systems combine SLAM and RGB-D geometry with object instances, scene graphs, voxel maps, and open-vocabulary features from CLIP or vision-language models; persistent fusion and active exploration support querying and navigation. For structured data, ontology and schema matching uses language-model embeddings, graph and instance evidence, constraints, retrieval, and human review to align concepts across sources. Current work is moving from static labels toward open-set, temporal, uncertainty-aware maps, while reliability under viewpoint or domain shifts, changing scenes, memory limits, and ambiguous mappings remains a central challenge.
Latest papers
93 papers11 papers in the last four weeks, down 15% on the four weeks before. 0.1% of all new papers.